Spanning tree protocol

Are there any techniques or guideline on how to assess a spanning tree protocol network?

You would have to look at the output from the switches management (show commands on a Cisco, Juniper, HP, etc.) to determine the version of STP running, the location of the root and which ports were blocking.

Also, is there a guideline or techniques on designing a LAN network to support users and servers?

The basic idea is to identify where the majority of the traffic is destined (or what traffic is the most important) and place the root so that traffic has the optimal path.

The problem is my client has a decent network size (100 Cisco switches) and the documentation is outdated. I can discover the LAN topology by using the Cisco commands. But that is a time consuming process. I am just wondering if there is any tools or techniques I can use.

There are some tools that can help. Solarwinds, Spiceworks, etc.

But if you're not familiar with them, there can be a long learning curve. And they're not always 100% accurate (the only thing worse than no documentation is bad documentation)

Nothing beats rolling up your sleeves and digging in. It's what we're paid for, right? ;-)

Rather than spend the time hunting down the logical topology of STP, why don't you just eliminate it?

Yeah, broadcast storms are so much fun. :-)

I said eliminate, not turn it off!

- "Eliminate"
Then how do you prevent loops?
- "not turn it off?
Not sure I understand. How do you eliminate STP without turn it off?

Well, how does STP work? To keep it simple lets assume you have a closet with 4 switches (3560G's) and a distribution switch in an MDF (4500). The four switches are daisy chained together (1 to 2, 2 to 3, 3 to 4). The top and bottom switch have uplinks back to the 4500. STP will shut one of those links down to keep a loop from occurring, right??

What if you disconnect all the switches from each other and then connect separate uplinks from each switch back to the distribution layer? You've just eliminated STP. Why? Because there's no loop, each switch is independent and therefore, STP does not shut down any links!

The idea is to try an design a well thought out network. Why design a network and allow a switch to shut a link off? It's a waste of a link. Design the network so you eliminate STP and all your links are active and running - you will end up increasing bandwidth - how?? By eliminating STP, you're taking 4 switches that were sharing a 3 Gig uplink and now you're able to do 4 Gig Ether-Channel from each switch. You went from 192 users over 3 links, to 48 users over 4 links x 4. Do the math!

Again, STP stops loops, but if you don't create them to begin with, STP won't drop the tree on any of your links!! :-)  Keep in mind, I'm not saying turn STP off, it's still on, you're just not giving it a reason to work!!! Now, if any yahoo goes in your closet and accidentally creates a loop, STP is still there to protect your network.

It works, this is how I design my networks, and they run great!

First off, your solution does not "eliminate" spanning-tree.  You even say, "I'm not saying turn STP off, it's still on". If spanning-tree is running, it hasn't been "eliminated".

It sounds like what you meant was to eliminate the need for spanning-tree. Which, in itself, is a good goal

Second, the author isn't asking how to redesign the network. They're saying that they need to asses the current design before redesigning the network. Which is a very good idea.

Pulling and reconnecting cables before knowing the current topology is a very bad idea.

DonJohnston -

Really?? This is what you're going to come at me with? I say potato you say patato!! So I should have said eliminate the "need" for STP and you would've understood what I was saying? Well than I guess you're right and I'm wrong.

You say, "Second, the author isn't asking how to redesign the network." But his question actually has two parts if you read it.

Mynet says, "If I were to redesign a LAN network with VLAN and STP, I need to assess the existing network to see where the root bridge is and what type of STP is deployed in the network." He saying he wants to redesign the network and he's trying to discover the STP topology for this purpose. I'm saying design the network so that STP doesn't come into play!!

My point was, if you're going to redesign the network, don't waste your time hunting down the operation of STP, just eliminate it - sorry, eliminate the "need" for it.

You also say "Pulling and reconnecting cables before knowing the current topology is a very bad idea." I never told Mynet to pull everything, unless you're speaking directly to my example, which was an example to help explain my thoughts.

Mynet -

To answer your other question, there are several show commands that will help you discover what is happening with STP in your network.

show spanning-tree
show spanning-tree detail
show spanning-tree summary
show spanning-tree interface [specify interface/vlan]

Now, the quick and painless way would be to use SNMP to discover the STP topology. There are a few avenues you can go down. 100 switches really isn't a lot, you could telnet into those and create text files of the STP topology in a few days. Good luck, I hope the sidebar between me and the other guy wasn't a distraction for you.

Really?? This is what you're going to come at me with?

Well, I'm not coming at you. Just pointing out there's a huge difference between "eliminating" and "eliminating the need for". And obviously I'm not the only one.

And you're correct, the author did ask if there was a guideline for designing a network.  And for that oversight, I apologize.

Apology accepted!

With regards to "eliminating" I look at it like this. If you have a newtork with loops in it and you get rid of those loops, what are you doing? I say eliminating them, you say?

It's a matter of words, my words confused you and others, so for that - I apologize!!

In the end, I only hope we were able to help mynet!!!

Good day gentlemen!

Well first off, is there any documentation at all (network diagrams, list of MAC addresses of the switches)?  If there is no documentation at all, then this could be a much more tedeous task.  if you do have the MAC addresses of the switches a good starting point would be to find the lowest MAC address  and check if it is the root or not and then keep moving down the list.  you could also use a network diagram and locate the most central switches and check those first.

If there are a lot of VLANs in your network and the previous network administrator did his job correctly you may have many root bridges.  

I would go through all the switches (yes it is time consuming) and issue the command show spanning-tree root this will show the bridge ID and the MAC of the Root bridge for that particular switch.  If the switch has many root bridges of different VLANs those should be shown also.  Then issue the show arp | inc xxxx.xxxx.xxxx replacing the X's with the MAC you just found and now you will have the IP address of the root bridge.  Hopefully you know the IPs of your switches.

As for design guidelines depends on how you want to set it up.  For 100 switches you would need at least a distribution layer with redundant switches connecting to the access layer where the hosts and servers connect to.  Depending on the expansion plans and expectations of your company/client you may want to implement a core layer so there will not need to be a redesign later.

If you are going to use STP then the root bridges should be located at the distribution layer for the VLANs that segment the servers and users. Also, you may want to use one distribution switch  as the bridge for VLANs 1, 2, and 3, while another switch is the bridge for VLANs 4, 5, and 6.  

here is a link to a cisco article on designing LAN.  You may know much of it already, or you may not.  In either case it is not a bad read and might give you some insight to how you want do design your network.

MAGO3,

I don't see the link. Thx

MAGO3,

I don't see the link. Thx